[go: up one dir, main page]

WO2022102105A1 - Conversion device, conversion method, and conversion program - Google Patents

Conversion device, conversion method, and conversion program Download PDF

Info

Publication number
WO2022102105A1
WO2022102105A1 PCT/JP2020/042528 JP2020042528W WO2022102105A1 WO 2022102105 A1 WO2022102105 A1 WO 2022102105A1 JP 2020042528 W JP2020042528 W JP 2020042528W WO 2022102105 A1 WO2022102105 A1 WO 2022102105A1
Authority
WO
WIPO (PCT)
Prior art keywords
subjective evaluation
evaluation value
conversion
audio signal
voice
Prior art date
Application number
PCT/JP2020/042528
Other languages
French (fr)
Japanese (ja)
Inventor
和徳 山田
航 光田
哲也 杵渕
裕司 青野
浩子 薮下
瑛彦 高島
孝 中村
Original Assignee
日本電信電話株式会社
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 日本電信電話株式会社 filed Critical 日本電信電話株式会社
Priority to PCT/JP2020/042528 priority Critical patent/WO2022102105A1/en
Priority to US18/036,598 priority patent/US20240013798A1/en
Priority to JP2022561229A priority patent/JPWO2022102105A1/ja
Publication of WO2022102105A1 publication Critical patent/WO2022102105A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/003—Changing voice quality, e.g. pitch or formants
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L21/00—Speech or voice signal processing techniques to produce another audible or non-audible signal, e.g. visual or tactile, in order to modify its quality or its intelligibility
    • G10L21/02—Speech enhancement, e.g. noise reduction or echo cancellation
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/48—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use
    • G10L25/51—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination
    • G10L25/60—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 specially adapted for particular use for comparison or discrimination for measuring the quality of voice signals

Definitions

  • the present invention relates to a conversion device, a conversion method, and a conversion program.
  • the conventional voice conversion method is a conversion targeting the explicit operation of parameters and the characteristics of the voice of the conversion destination, so the conversion is not always subjectively easy for the listener to hear.
  • the present invention has been made in view of the above, and an object of the present invention is to provide a conversion device, a conversion method, and a conversion program capable of converting an input voice into a voice that is subjectively easy for the listener to hear. do.
  • the conversion device is any value of the subjective evaluation value which quantifies the ease of transmitting the content of the voice perceived by a person from the input voice signal. It has an evaluation unit that estimates whether or not to take, and a conversion unit that converts an input voice signal so that the subjective evaluation value becomes a predetermined value based on the subjective evaluation value estimated by the evaluation unit. It is characterized by.
  • the input voice can be converted into a voice that is subjectively easy for the listener to hear.
  • FIG. 1 is a diagram showing an example of the configuration of the conversion device according to the first embodiment.
  • FIG. 2 is a flowchart showing a processing procedure of the conversion processing according to the first embodiment.
  • FIG. 3 is a diagram showing an example of the configuration of the conversion device according to the second embodiment.
  • FIG. 4 is a flowchart showing a processing procedure of the conversion processing according to the second embodiment.
  • FIG. 5 is a diagram showing an example of a computer in which a conversion device is realized by executing a program.
  • the conversion device according to the first embodiment converts a voice signal by utilizing the subjective evaluation tendency for voice.
  • the conversion device according to the first embodiment converts the input voice based on the subjective evaluation value that quantifies the ease of transmitting the content of the voice felt by a person, so that, for example, it is easy for the listener to hear subjectively. It is converted to the voice.
  • FIG. 1 is a diagram showing an example of the configuration of the conversion device according to the first embodiment.
  • a predetermined program is read into a computer or the like including a ROM (Read Only Memory), a RAM (Random Access Memory), a CPU (Central Processing Unit), and the CPU is predetermined. It is realized by executing the program of.
  • the conversion device 10 has an evaluation unit 11 and a conversion unit 12.
  • a speaker's voice signal is input to the conversion device 10.
  • the conversion device 10 converts the input audio signal into, for example, an audio signal that is subjectively easy for the listener to hear and outputs the signal.
  • the evaluation unit 11 estimates which of the subjective evaluation values should be taken from the input audio signal.
  • the subjective evaluation value is a numerical value of the ease with which the content of the voice felt by a person is transmitted.
  • the subjective evaluation value is, for example, a numerical value indicating items such as ease of understanding, naturalness of voice, ease of understanding of content, appropriateness of spacing, goodness of speaking, or degree of impression. Is.
  • the subjective evaluation value is, for example, one person or a plurality of people evaluating the audio signal in N stages (for example, 5 stages) for each item, and the evaluation value evaluated for a plurality of subjective evaluation items. Is represented by a vector.
  • the subjective evaluation value if it is a subjective evaluation value by a plurality of people, for example, a value obtained by averaging the subjective evaluation values of a plurality of people is used for each item.
  • the evaluation unit 11 extracts a feature amount from the input audio signal, and estimates a subjective evaluation value using an evaluation model based on the extracted feature amount.
  • the evaluation model is a model in which the relationship between the feature amount of the audio signal for learning and the subjective evaluation value corresponding to the audio signal for learning is learned.
  • the evaluation model learns the relationship between the feature amount of a plurality of learning voice signals to which a subjective evaluation value is given for each item and the subjective evaluation value, for example, by using a regression method using machine learning. As a result, the evaluation model estimates the subjective evaluation value based on the feature amount extracted from the input audio signal.
  • the conversion unit 12 converts the input audio signal so as to have a predetermined subjective evaluation value based on the subjective evaluation value estimated by the evaluation unit 11. For example, the conversion unit 12 sets the upper limit of the subjective evaluation value as a fixed value in advance as a predetermined value, and converts the input audio signal so as to be the upper limit of the subjective evaluation value.
  • the conversion unit 12 extracts the feature amount from the input audio signal. Then, the conversion unit 12 converts the input audio signal into a subjective evaluation value of a predetermined value by using the conversion model based on the extracted feature amount.
  • the conversion model is a model that learns the conversion from the feature amount of the input audio signal to the feature amount of the audio signal which is a subjective evaluation value of a predetermined value.
  • the conversion unit 12 inputs the feature amount of the audio signal and the subjective evaluation value of the audio signal into the conversion model, and outputs the feature amount of the audio signal which is a predetermined subjective evaluation value. To get. Then, the conversion unit 12 converts the acquired feature amount into an audio signal to obtain an audio signal which is a predetermined subjective evaluation value.
  • the conversion unit 12 outputs the acquired audio signal to the outside as an output of the conversion device 10.
  • the learning of this transformation model will be explained.
  • a plurality of audio signals that speak the same content and subjective evaluation values corresponding to each audio signal are used as learning data.
  • These learning data have the same audio content, but differ in subjective evaluation values (naturalness, comprehensibility, etc.).
  • These learning data are, for example, feature quantities of audio signals to which subjective evaluation values of 1 to 5 are given as learning data.
  • the conversion model is based on, for example, the difference between the subjective evaluation value of the easy-to-understand item being 1 (first subjective evaluation value) and the subjective evaluation value of the easy-to-understand item being 5 (second subjective evaluation value). Learn to convert audio signal features.
  • a feature amount of a voice signal having a bad subjective evaluation value (a voice signal to which a first subjective evaluation value is given) is used as an input of a conversion model, and a voice signal having a good subjective evaluation value (a value different from the first subjective evaluation value) is used as an input.
  • the input / output relationship is learned by using, for example, machine learning, using the feature amount of the second subjective evaluation value (speech signal) as an output, and is used as a conversion model.
  • the subjective evaluation values of the audio signals on the output side and the input side are specifically used as auxiliary inputs.
  • the difference vector between the two is used as the auxiliary input.
  • a conversion model in which the input / output relationship and the subjective evaluation value (difference) are associated with each other can be obtained by learning.
  • FIG. 2 is a flowchart showing a processing procedure of the conversion processing according to the first embodiment.
  • the evaluation unit 11 estimates which of the subjective evaluation values is to be taken from the input audio signal. Processing is performed (step S2), and the subjective evaluation value for the input audio signal is output (step S3).
  • the conversion unit 12 converts the input audio signal so as to have a predetermined subjective evaluation value based on the subjective evaluation value estimated by the evaluation unit 11 (step S4), and the converted audio.
  • a signal is output (step S5).
  • the subjective evaluation value is estimated from the input audio signal, and the subjective evaluation value of a predetermined value is obtained based on the estimated subjective evaluation value. Converts the input audio signal to.
  • the subjective evaluation value is a numerical value of how easy it is to convey the content of the voice that a person feels. It is a step-by-step evaluation of the goodness of speaking or the degree of impression.
  • the input audio signal is converted, for example, so that the subjective evaluation value becomes the upper limit value based on the subjective evaluation value shown above estimated from the input audio signal. Therefore, according to the first embodiment, by utilizing not only the objective characteristics or physical characteristics of the audio signal but also the subjective evaluation value of the listener, the listener can subjectively easily hear the natural audio. It can be converted into a signal.
  • an evaluation model for estimating the subjective evaluation value of the input audio signal by learning the correspondence relationship between the audio signal and the subjective evaluation value, a plurality of audio signals, and each audio signal A conversion model is used in which the input audio signal is converted into an audio signal which is a predetermined subjective evaluation value by learning the subjective evaluation value of. Therefore, in the first embodiment, the listener subjectively listens to the input audio signal according to the characteristics by utilizing the correspondence between the audio signal and the subjective evaluation value for the evaluation and conversion of the audio signal. It can be appropriately converted into an easy audio signal.
  • FIG. 3 is a diagram showing an example of the configuration of the conversion device according to the second embodiment.
  • the conversion device 210 As shown in FIG. 3, the conversion device 210 according to the second embodiment has a conversion unit 212 instead of the conversion unit 12 shown in FIG. Further, the conversion device 210 accepts the input of the voice signal to be converted and also receives the input of the evaluation information of the target voice.
  • the evaluation information of the target voice is a subjective evaluation value that is the target of the converted voice signal.
  • a target value is set for each item.
  • the predetermined subjective evaluation value was a fixed value in the first embodiment, but in the second embodiment, it becomes a target subjective evaluation value input from the outside.
  • the conversion unit 212 converts the input audio signal so that the subjective evaluation value estimated by the evaluation unit 11 becomes the target subjective evaluation value.
  • the conversion unit 212 converts the input audio signal so as to have a target subjective evaluation value input from the outside (for example, a listener or a speaker). This target subjective evaluation value may be input as evaluation information whose target voice is how much the speaker himself / herself wants to improve his / her voice.
  • the conversion unit 212 extracts the feature amount from the input audio signal. Then, the conversion unit 212 converts the input audio signal into the target subjective evaluation value by using the conversion model based on the extracted feature amount.
  • the conversion model is a model that learns the conversion from the feature amount of the input audio signal to the feature amount of the audio signal which is the target subjective evaluation value.
  • the conversion unit 212 inputs the feature amount of the audio signal and the subjective evaluation value of the audio signal into the conversion model, so that the converted audio signal which is the target subjective evaluation value can be obtained. Obtain the output of the feature quantity.
  • the conversion unit 212 converts the acquired feature amount into an audio signal to obtain an audio signal which is a target subjective evaluation value.
  • the conversion unit 212 outputs the acquired audio signal to the outside as an output of the conversion device 210.
  • the learning of the transformation model may be performed in the same manner as in the first embodiment.
  • FIG. 4 is a flowchart showing a processing procedure of the conversion processing according to the second embodiment.
  • Steps S11 to S13 shown in FIG. 3 are the same processes as steps S1 to S3 shown in FIG.
  • the conversion device 210 receives the input of the evaluation information of the target voice (step S14)
  • the conversion device 210 sets the target subjective evaluation value indicated by the evaluation information of the target voice based on the subjective evaluation value estimated by the evaluation unit 11.
  • the input audio signal is converted (step S15), and the converted audio signal is output (step S16).
  • the input audio signal is converted so that the subjective evaluation value of the audio signal estimated by the evaluation unit 11 becomes the target subjective evaluation value.
  • the subjective evaluation value after conversion is fixed. If the subjective evaluation value after conversion is fixed as in the first embodiment, it may not be possible to deal with flexible and complicated conversion according to various situations and listeners.
  • the second embodiment by explicitly inputting the subjective evaluation value of the target, it is possible to flexibly cope with a complicated case where the desired voice is set at a different stage for each item. And can be converted into an audio signal that suits the listener.
  • each component of each of the illustrated devices is functional and conceptual, and does not necessarily have to be physically configured as shown in the figure. That is, the specific form of distribution / integration of each device is not limited to the one shown in the figure, and all or part of them may be functionally or physically distributed / physically in arbitrary units according to various loads and usage conditions. Can be integrated and configured.
  • the conversion devices 10 and 210 may be an integrated device.
  • each processing function performed by each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by wired logic.
  • all or part of the processes described as being automatically performed can be manually performed, or the processes described as being manually performed can be performed. All or part of it can be done automatically by a known method.
  • the processes described in the present embodiment are not only executed in chronological order according to the order of description, but may also be executed in parallel or individually as required by the processing capacity of the device that executes the processes. ..
  • the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above document and drawings can be arbitrarily changed unless otherwise specified.
  • FIG. 5 is a diagram showing an example of a computer in which the conversion devices 10 and 210 are realized by executing the program.
  • the computer 1000 has, for example, a memory 1010 and a CPU 1020.
  • the computer 1000 also has a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. Each of these parts is connected by a bus 1080.
  • Memory 1010 includes ROM 1011 and RAM 1012.
  • the ROM 1011 stores, for example, a boot program such as a BIOS (Basic Input Output System).
  • BIOS Basic Input Output System
  • the hard disk drive interface 1030 is connected to the hard disk drive 1031.
  • the disk drive interface 1040 is connected to the disk drive 1041.
  • a removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1041.
  • the serial port interface 1050 is connected to, for example, a mouse 1110 and a keyboard 1120.
  • the video adapter 1060 is connected to, for example, the display 1130.
  • the hard disk drive 1031 stores, for example, the OS 1091, the application program 1092, the program module 1093, and the program data 1094. That is, the program that defines each process of the conversion devices 10 and 210 is implemented as a program module 1093 in which a code that can be executed by the computer 1000 is described.
  • the program module 1093 is stored in, for example, the hard disk drive 1031.
  • the program module 1093 for executing the same processing as the functional configuration in the conversion devices 10 and 210 is stored in the hard disk drive 1031.
  • the hard disk drive 1031 may be replaced by an SSD (Solid State Drive).
  • the setting data used in the processing of the above-described embodiment is stored as program data 1094 in, for example, the memory 1010 or the hard disk drive 1031. Then, the CPU 1020 reads the program module 1093 and the program data 1094 stored in the memory 1010 and the hard disk drive 1031 into the RAM 1012 and executes them as needed.
  • the program module 1093 and the program data 1094 are not limited to those stored in the hard disk drive 1031 but may be stored in, for example, a removable storage medium and read by the CPU 1020 via the disk drive 1041 or the like. Alternatively, the program module 1093 and the program data 1094 may be stored in another computer connected via a network (LAN (Local Area Network), WAN (Wide Area Network), etc.). Then, the program module 1093 and the program data 1094 may be read from another computer by the CPU 1020 via the network interface 1070. Further, the processing of the neural network used in the conversion devices 10, 210 and the learning devices 20, 220, 320, 420 may be executed by using the GPU.
  • LAN Local Area Network
  • WAN Wide Area Network

Landscapes

  • Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Computational Linguistics (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • Acoustics & Sound (AREA)
  • Multimedia (AREA)
  • User Interface Of Digital Computer (AREA)
  • Telephonic Communication Services (AREA)

Abstract

This conversion device (10) has: an evaluation unit (11) that infers, from an inputted audio signal, which values to adopt among subjective evaluation values that quantify the ease with which a listener experiences the conveying of audio content; and a conversion unit (12) that converts the inputted audio signal so as to reach a prescribed subjective evaluation value on the basis of the subjective evaluation values inferred by the evaluation unit (11).

Description

変換装置、変換方法及び変換プログラムConversion device, conversion method and conversion program
　本発明は、変換装置、変換方法及び変換プログラムに関する。 The present invention relates to a conversion device, a conversion method, and a conversion program.
　従来、音声の周波数成分や話速などの特徴を変更し、別の声質の音声に変換を行う音声変換方法が提案されている（例えば、特許文献１参照）。 Conventionally, a voice conversion method has been proposed in which characteristics such as the frequency component and speaking speed of voice are changed to convert to voice with a different voice quality (see, for example, Patent Document 1).
特許第２６１２８６９号公報Japanese Patent No. 2612869
　従来の音声変換方法は、パラメータの明示的な操作や、変換先の音声の特徴を目標とした変換であるため、必ずしも聞き手に対して主観的に聞きやすく変換されるわけではない。 The conventional voice conversion method is a conversion targeting the explicit operation of parameters and the characteristics of the voice of the conversion destination, so the conversion is not always subjectively easy for the listener to hear.
　本発明は、上記に鑑みてなされたものであって、入力音声を、聞き手に対して主観的に聞きやすくなる音声に変換することができる変換装置、変換方法及び変換プログラム提供することを目的とする。 The present invention has been made in view of the above, and an object of the present invention is to provide a conversion device, a conversion method, and a conversion program capable of converting an input voice into a voice that is subjectively easy for the listener to hear. do.
　上述した課題を解決し、目的を達成するために、本発明に係る変換装置は、入力された音声信号から、人が感じる音声の内容の伝わりやすさを数値化した主観評価値のいずれの値を取るかを推定する評価部と、評価部によって推定された主観評価値を基に、所定の値の主観評価値となるように、入力された音声信号を変換する変換部と、を有することを特徴とする。 In order to solve the above-mentioned problems and achieve the object, the conversion device according to the present invention is any value of the subjective evaluation value which quantifies the ease of transmitting the content of the voice perceived by a person from the input voice signal. It has an evaluation unit that estimates whether or not to take, and a conversion unit that converts an input voice signal so that the subjective evaluation value becomes a predetermined value based on the subjective evaluation value estimated by the evaluation unit. It is characterized by.
　本発明によれば、入力音声を、聞き手に対して主観的に聞きやすくなる音声に変換することができる。 According to the present invention, the input voice can be converted into a voice that is subjectively easy for the listener to hear.
図１は、実施の形態１に係る変換装置の構成の一例を示す図である。FIG. 1 is a diagram showing an example of the configuration of the conversion device according to the first embodiment. 図２は、実施の形態１に係る変換処理の処理手順を示すフローチャートである。FIG. 2 is a flowchart showing a processing procedure of the conversion processing according to the first embodiment. 図３は、実施の形態２に係る変換装置の構成の一例を示す図である。FIG. 3 is a diagram showing an example of the configuration of the conversion device according to the second embodiment. 図４は、実施の形態２に係る変換処理の処理手順を示すフローチャートである。FIG. 4 is a flowchart showing a processing procedure of the conversion processing according to the second embodiment. 図５は、プログラムが実行されることにより、変換装置が実現されるコンピュータの一例を示す図である。FIG. 5 is a diagram showing an example of a computer in which a conversion device is realized by executing a program.
　以下に、本願に係る変換装置、変換方法及び変換プログラムの実施形態を図面に基づいて詳細に説明する。なお、本発明は、以下に説明する実施形態により限定されるものではない。 Hereinafter, embodiments of the conversion device, conversion method, and conversion program according to the present application will be described in detail with reference to the drawings. The present invention is not limited to the embodiments described below.
［実施の形態１］
［変換装置］
　まず、実施の形態１に係る変換装置について説明する。実施の形態１に係る変換装置は、音声に対する主観評価傾向を利用して、音声信号を変換する。実施の形態１に係る変換装置は、人が感じる音声の内容の伝わりやすさを数値化した主観評価値を基に、入力音声を変換することで、例えば、聞き手に対して主観的に聞きやすくなる音声に変換している。 [Embodiment 1]
[Converter]
First, the conversion device according to the first embodiment will be described. The conversion device according to the first embodiment converts a voice signal by utilizing the subjective evaluation tendency for voice. The conversion device according to the first embodiment converts the input voice based on the subjective evaluation value that quantifies the ease of transmitting the content of the voice felt by a person, so that, for example, it is easy for the listener to hear subjectively. It is converted to the voice.
［変換装置］
　図１は、実施の形態１に係る変換装置の構成の一例を示す図である。実施の形態１に係る変換装置１０は、例えば、ＲＯＭ（Read　Only　Memory）、ＲＡＭ（Random　Access　Memory）、ＣＰＵ（Central　Processing　Unit）等を含むコンピュータ等に所定のプログラムが読み込まれて、ＣＰＵが所定のプログラムを実行することで実現される。 [Converter]
FIG. 1 is a diagram showing an example of the configuration of the conversion device according to the first embodiment. In the conversion device 10 according to the first embodiment, for example, a predetermined program is read into a computer or the like including a ROM (Read Only Memory), a RAM (Random Access Memory), a CPU (Central Processing Unit), and the CPU is predetermined. It is realized by executing the program of.
　図１に示すように、変換装置１０は、評価部１１及び変換部１２を有する。変換装置１０には、話し手の音声信号が入力される。変換装置１０は、入力された音声信号を、例えば、聞き手に対して主観的に聞きやすくなる音声信号に変換して出力する。 As shown in FIG. 1, the conversion device 10 has an evaluation unit 11 and a conversion unit 12. A speaker's voice signal is input to the conversion device 10. The conversion device 10 converts the input audio signal into, for example, an audio signal that is subjectively easy for the listener to hear and outputs the signal.
　評価部１１は、入力された音声信号から、主観評価値のいずれの値を取るかを推定する。ここで、主観評価値は、人が感じる音声の内容の伝わりやすさを数値化したものである。 The evaluation unit 11 estimates which of the subjective evaluation values should be taken from the input audio signal. Here, the subjective evaluation value is a numerical value of the ease with which the content of the voice felt by a person is transmitted.
　主観評価値は、例えば、理解のしやすさ、音声の自然さ、内容の分かりやすさ、間の取り方の適切さ、話し方のうまさ、または、印象度合いの項目を、それぞれ数値で示したものである。主観評価値は、例えば、一人の人、または、複数の人が、音声信号を、各項目についてＮ段階（例えば、５段階）で評価したものであり、複数の主観評価項目について評価した評価値は、ベクトルで表現される。主観評価値は、複数の人による主観評価値であれば、例えば、項目ごとに複数人の主観評価値を平均した値を用いる。 The subjective evaluation value is, for example, a numerical value indicating items such as ease of understanding, naturalness of voice, ease of understanding of content, appropriateness of spacing, goodness of speaking, or degree of impression. Is. The subjective evaluation value is, for example, one person or a plurality of people evaluating the audio signal in N stages (for example, 5 stages) for each item, and the evaluation value evaluated for a plurality of subjective evaluation items. Is represented by a vector. As the subjective evaluation value, if it is a subjective evaluation value by a plurality of people, for example, a value obtained by averaging the subjective evaluation values of a plurality of people is used for each item.
　評価部１１は、入力された音声信号から特徴量を抽出し、抽出した特徴量を基に、評価モデルを用いて、主観評価値を推定する。評価モデルは、学習用の音声信号の特徴量と、この学習用の音声信号に対応する主観評価値との関係性を学習したモデルである。 The evaluation unit 11 extracts a feature amount from the input audio signal, and estimates a subjective evaluation value using an evaluation model based on the extracted feature amount. The evaluation model is a model in which the relationship between the feature amount of the audio signal for learning and the subjective evaluation value corresponding to the audio signal for learning is learned.
　評価モデルは、項目ごとに主観評価値が付与された複数の学習用音声信号の特徴量と主観評価値との関係を、例えば、機械学習を用いた回帰による手法を用いて学習する。これによって、評価モデルは、入力された音声信号から抽出した特徴量を基に、主観評価値を推定する。 The evaluation model learns the relationship between the feature amount of a plurality of learning voice signals to which a subjective evaluation value is given for each item and the subjective evaluation value, for example, by using a regression method using machine learning. As a result, the evaluation model estimates the subjective evaluation value based on the feature amount extracted from the input audio signal.
　変換部１２は、評価部１１によって推定された主観評価値を基に、所定の値の主観評価値となるように、入力された音声信号を変換する。変換部１２は、例えば、所定の値として、主観評価値の上限値をあらかじめ固定値として設定し、主観評価値の上限値となるように、入力された音声信号を変換する。 The conversion unit 12 converts the input audio signal so as to have a predetermined subjective evaluation value based on the subjective evaluation value estimated by the evaluation unit 11. For example, the conversion unit 12 sets the upper limit of the subjective evaluation value as a fixed value in advance as a predetermined value, and converts the input audio signal so as to be the upper limit of the subjective evaluation value.
　変換部１２は、入力された音声信号から特徴量を抽出する。そして、変換部１２は、抽出した特徴量を基に、変換モデルを用いて、入力された音声信号を、所定の値の主観評価値となるように変換する。変換モデルは、入力された音声信号の特徴量から、所定の値の主観評価値である音声信号の特徴量への変換を学習したモデルである。変換部１２は、変換の際には、変換モデルに、音声信号の特徴量と、その音声信号の主観評価値とを入力することで、所定の主観評価値である音声信号の特徴量の出力を取得する。そして、変換部１２は、取得した特徴量を音声信号に変換することで、所定の主観評価値である音声信号を得る。変換部１２は、取得した音声信号を、変換装置１０の出力として、外部に出力する。 The conversion unit 12 extracts the feature amount from the input audio signal. Then, the conversion unit 12 converts the input audio signal into a subjective evaluation value of a predetermined value by using the conversion model based on the extracted feature amount. The conversion model is a model that learns the conversion from the feature amount of the input audio signal to the feature amount of the audio signal which is a subjective evaluation value of a predetermined value. At the time of conversion, the conversion unit 12 inputs the feature amount of the audio signal and the subjective evaluation value of the audio signal into the conversion model, and outputs the feature amount of the audio signal which is a predetermined subjective evaluation value. To get. Then, the conversion unit 12 converts the acquired feature amount into an audio signal to obtain an audio signal which is a predetermined subjective evaluation value. The conversion unit 12 outputs the acquired audio signal to the outside as an output of the conversion device 10.
　この変換モデルの学習について説明する。まず、同一の内容を話した複数の音声信号と、各音声信号に対応する主観評価値とを学習用データとする。これらの学習用データは、音声の内容は同一だが、主観評価値（自然さや分かりやすさ等）に差があるものである。これらの学習用データは、例えば、学習用データとして、１～５の主観評価値が付与された音声信号の特徴量である。変換モデルは、例えば、分かりやすさの項目の主観評価値が１（第１の主観評価値）と、分かりやすさの項目の主観評価値が５（第２の主観評価値）との違いによる音声信号の特徴量の変換を学習する。例えば、主観評価値の悪い音声信号（第１の主観評価値が付与された音声信号）の特徴量を変換モデルの入力とし、主観評価値の良い音声信号（第１の主観評価値と異なる値である第２の主観評価値が付与された音声信号）の特徴量を出力として、入出力関係を、例えば機械学習を用いて学習し、変換モデルとする。 The learning of this transformation model will be explained. First, a plurality of audio signals that speak the same content and subjective evaluation values corresponding to each audio signal are used as learning data. These learning data have the same audio content, but differ in subjective evaluation values (naturalness, comprehensibility, etc.). These learning data are, for example, feature quantities of audio signals to which subjective evaluation values of 1 to 5 are given as learning data. The conversion model is based on, for example, the difference between the subjective evaluation value of the easy-to-understand item being 1 (first subjective evaluation value) and the subjective evaluation value of the easy-to-understand item being 5 (second subjective evaluation value). Learn to convert audio signal features. For example, a feature amount of a voice signal having a bad subjective evaluation value (a voice signal to which a first subjective evaluation value is given) is used as an input of a conversion model, and a voice signal having a good subjective evaluation value (a value different from the first subjective evaluation value) is used as an input. The input / output relationship is learned by using, for example, machine learning, using the feature amount of the second subjective evaluation value (speech signal) as an output, and is used as a conversion model.
　変換モデルの学習の際、出力側と入力側の音声信号の主観評価値は、具体的には、補助の入力として用いる。例えば、補助の入力として、両者の差ベクトル（出力側の主観評価値－入力側の主観評価値）を用いる。これによって、入出力関係と主観評価値と（の差）を対応付けた変換モデルを学習によって得ることができる。 When learning the transformation model, the subjective evaluation values of the audio signals on the output side and the input side are specifically used as auxiliary inputs. For example, the difference vector between the two (subjective evaluation value on the output side-subjective evaluation value on the input side) is used as the auxiliary input. As a result, a conversion model in which the input / output relationship and the subjective evaluation value (difference) are associated with each other can be obtained by learning.
［変換処理の処理手順］
　次に、変換装置１０における変換処理について説明する。図２は、実施の形態１に係る変換処理の処理手順を示すフローチャートである。 [Conversion processing procedure]
Next, the conversion process in the conversion device 10 will be described. FIG. 2 is a flowchart showing a processing procedure of the conversion processing according to the first embodiment.
　図２に示すように、変換装置１０は、音声信号の入力を受け付けると（ステップＳ１）、評価部１１は、入力された音声信号から、主観評価値のいずれの値を取るかを推定する評価処理を行い（ステップＳ２）、入力された音声信号に対する主観評価値を出力する（ステップＳ３）。 As shown in FIG. 2, when the conversion device 10 receives the input of the audio signal (step S1), the evaluation unit 11 estimates which of the subjective evaluation values is to be taken from the input audio signal. Processing is performed (step S2), and the subjective evaluation value for the input audio signal is output (step S3).
　そして、変換部１２は、評価部１１によって推定された主観評価値を基に、所定の値の主観評価値となるように、入力された音声信号を変換し（ステップＳ４）、変換後の音声信号を出力する（ステップＳ５）。 Then, the conversion unit 12 converts the input audio signal so as to have a predetermined subjective evaluation value based on the subjective evaluation value estimated by the evaluation unit 11 (step S4), and the converted audio. A signal is output (step S5).
［実施の形態１の効果］
　このように、実施の形態１では、入力された音声信号から、主観評価値のいずれの値を取るかを推定し、推定した主観評価値を基に、所定の値の主観評価値となるように、入力された音声信号を変換する。主観評価値は、人が感じる音声の内容の伝わりやすさを数値化したものであり、例えば、理解のしやすさ、音声の自然さ、内容の分かりやすさ、間の取り方の適切さ、話し方のうまさ、または、印象度合いを、段階的に評価したものである。 [Effect of Embodiment 1]
As described above, in the first embodiment, which value of the subjective evaluation value is estimated from the input audio signal, and the subjective evaluation value of a predetermined value is obtained based on the estimated subjective evaluation value. Converts the input audio signal to. The subjective evaluation value is a numerical value of how easy it is to convey the content of the voice that a person feels. It is a step-by-step evaluation of the goodness of speaking or the degree of impression.
　実施の形態１では、入力された音声信号から推定した、上記に示す主観評価値を基に、例えば、主観評価値が上限値となるように、入力された音声信号を変換する。このため、実施の形態１によれば、音声信号の客観的特徴或いは物理的特徴だけではなく、聞く人間の主観的な評価値を活用することで、聞き手が主観的に聞きやすく、自然な音声信号に変換することができる。 In the first embodiment, the input audio signal is converted, for example, so that the subjective evaluation value becomes the upper limit value based on the subjective evaluation value shown above estimated from the input audio signal. Therefore, according to the first embodiment, by utilizing not only the objective characteristics or physical characteristics of the audio signal but also the subjective evaluation value of the listener, the listener can subjectively easily hear the natural audio. It can be converted into a signal.
　そして、実施の形態１では、音声信号と、その主観評価値との対応関係を学習させることで入力された音声信号の主観評価値を推定する評価モデルと、複数の音声信号と、各音声信号の主観評価値とを学習させて、入力された音声信号を所定の主観評価値である音声信号に変換する変換モデルとを用いる。したがって、実施の形態１では、音声信号と主観評価値との対応関係を、音声信号の評価及び変換に活用することで、入力された音声信号を、特徴に応じて、聞き手が主観的に聞きやすい音声信号に適切に変換することができる。 Then, in the first embodiment, an evaluation model for estimating the subjective evaluation value of the input audio signal by learning the correspondence relationship between the audio signal and the subjective evaluation value, a plurality of audio signals, and each audio signal. A conversion model is used in which the input audio signal is converted into an audio signal which is a predetermined subjective evaluation value by learning the subjective evaluation value of. Therefore, in the first embodiment, the listener subjectively listens to the input audio signal according to the characteristics by utilizing the correspondence between the audio signal and the subjective evaluation value for the evaluation and conversion of the audio signal. It can be appropriately converted into an easy audio signal.
［実施の形態２］
　次に、実施の形態２について説明する。図３は、実施の形態２に係る変換装置の構成の一例を示す図である。 [Embodiment 2]
Next, the second embodiment will be described. FIG. 3 is a diagram showing an example of the configuration of the conversion device according to the second embodiment.
　図３に示すように、実施の形態２に係る変換装置２１０は、図１に示す変換部１２に代えて、変換部２１２を有する。また、変換装置２１０は、変換対象の音声信号の入力を受け付けるとともに、目標音声の評価情報の入力を受け付ける。目標音声の評価情報は、具体的には、変換後の音声信号の目標とする主観評価値である。また、主観評価値が複数の項目を有する場合には、項目ごとに目標値が設定される。なお、所定の主観評価値は、実施の形態１において固定値であったが、実施の形態２では、外部から入力された目標とする主観評価値となる。 As shown in FIG. 3, the conversion device 210 according to the second embodiment has a conversion unit 212 instead of the conversion unit 12 shown in FIG. Further, the conversion device 210 accepts the input of the voice signal to be converted and also receives the input of the evaluation information of the target voice. Specifically, the evaluation information of the target voice is a subjective evaluation value that is the target of the converted voice signal. When the subjective evaluation value has a plurality of items, a target value is set for each item. The predetermined subjective evaluation value was a fixed value in the first embodiment, but in the second embodiment, it becomes a target subjective evaluation value input from the outside.
　変換部２１２は、評価部１１によって推定された主観評価値が、目標とする主観評価値となるように、入力された音声信号を変換する。変換部２１２は、外部（例えば、聞き手や話し手）から入力された目標とする主観評価値となるように、入力された音声信号を変換する。この目標とする主観評価値は、話し手本人がどのくらい自分の音声を良くしたいかを目標音声とする評価情報として入力する場合もある。 The conversion unit 212 converts the input audio signal so that the subjective evaluation value estimated by the evaluation unit 11 becomes the target subjective evaluation value. The conversion unit 212 converts the input audio signal so as to have a target subjective evaluation value input from the outside (for example, a listener or a speaker). This target subjective evaluation value may be input as evaluation information whose target voice is how much the speaker himself / herself wants to improve his / her voice.
　変換部２１２は、入力された音声信号から特徴量を抽出する。そして、変換部２１２は、抽出した特徴量を基に、変換モデルを用いて、入力された音声信号を、目標とする主観評価値となるように変換する。変換モデルは、入力された音声信号の特徴量から、目標とする主観評価値である音声信号の特徴量への変換を学習したモデルである。変換部２１２は、変換の際には、変換モデルに、音声信号の特徴量と、その音声信号の主観評価値とを入力することで、目標とする主観評価値である変換された音声信号の特徴量の出力を得る。そして、変換部２１２は、取得した特徴量を音声信号に変換することで、目標とする主観評価値である音声信号を得る。変換部２１２は、取得した音声信号を、変換装置２１０の出力として、外部に出力する。なお、変換モデルの学習は、実施の形態１と同様に行えばよい。 The conversion unit 212 extracts the feature amount from the input audio signal. Then, the conversion unit 212 converts the input audio signal into the target subjective evaluation value by using the conversion model based on the extracted feature amount. The conversion model is a model that learns the conversion from the feature amount of the input audio signal to the feature amount of the audio signal which is the target subjective evaluation value. At the time of conversion, the conversion unit 212 inputs the feature amount of the audio signal and the subjective evaluation value of the audio signal into the conversion model, so that the converted audio signal which is the target subjective evaluation value can be obtained. Obtain the output of the feature quantity. Then, the conversion unit 212 converts the acquired feature amount into an audio signal to obtain an audio signal which is a target subjective evaluation value. The conversion unit 212 outputs the acquired audio signal to the outside as an output of the conversion device 210. The learning of the transformation model may be performed in the same manner as in the first embodiment.
［変換処理の処理手順］
　次に、変換装置２１０における変換処理について説明する。図４は、実施の形態２に係る変換処理の処理手順を示すフローチャートである。 [Conversion processing procedure]
Next, the conversion process in the conversion device 210 will be described. FIG. 4 is a flowchart showing a processing procedure of the conversion processing according to the second embodiment.
　図３に示すステップＳ１１～ステップＳ１３は、図２に示すステップＳ１～Ｓ３と同じ処理である。変換装置２１０は、目標音声の評価情報の入力を受け付けると（ステップＳ１４）、評価部１１によって推定された主観評価値を基に、目標音声の評価情報で示された目標とする主観評価値となるように、入力された音声信号を変換し（ステップＳ１５）、変換後の音声信号を出力する（ステップＳ１６）。 Steps S11 to S13 shown in FIG. 3 are the same processes as steps S1 to S3 shown in FIG. When the conversion device 210 receives the input of the evaluation information of the target voice (step S14), the conversion device 210 sets the target subjective evaluation value indicated by the evaluation information of the target voice based on the subjective evaluation value estimated by the evaluation unit 11. The input audio signal is converted (step S15), and the converted audio signal is output (step S16).
［実施の形態２の効果］
　このように、実施の形態２では、評価部１１によって推定された音声信号の主観評価値が、目標とする主観評価値となるように、入力された音声信号を変換する。ここで、実施の形態１では、変換後の主観評価値が固定されている例を説明した。この実施の形態１のように、変換後の主観評価値が固定されていると、多様な状況や聞き手に応じた柔軟かつ複雑な変換に対応できない場合が考えられる。 [Effect of Embodiment 2]
As described above, in the second embodiment, the input audio signal is converted so that the subjective evaluation value of the audio signal estimated by the evaluation unit 11 becomes the target subjective evaluation value. Here, in the first embodiment, an example in which the subjective evaluation value after conversion is fixed has been described. If the subjective evaluation value after conversion is fixed as in the first embodiment, it may not be possible to deal with flexible and complicated conversion according to various situations and listeners.
　これに対し、実施の形態２では、目標の主観評価値を明示的に入力できるようにすることで、所望の音声が各項目で異なる段階に設定される複雑な場合にも柔軟に対応することができ、聞き手に合った音声信号に変換することができる。 On the other hand, in the second embodiment, by explicitly inputting the subjective evaluation value of the target, it is possible to flexibly cope with a complicated case where the desired voice is set at a different stage for each item. And can be converted into an audio signal that suits the listener.
［システム構成等］
　図示した各装置の各構成要素は機能概念的なものであり、必ずしも物理的に図示の如く構成されていることを要しない。すなわち、各装置の分散・統合の具体的形態は図示のものに限られず、その全部又は一部を、各種の負荷や使用状況等に応じて、任意の単位で機能的又は物理的に分散・統合して構成することができる。例えば、変換装置１０，２１０は、一体の装置であってもよい。さらに、各装置にて行なわれる各処理機能は、その全部又は任意の一部が、ＣＰＵ及び当該ＣＰＵにて解析実行されるプログラムにて実現され、あるいは、ワイヤードロジックによるハードウェアとして実現され得る。 [System configuration, etc.]
Each component of each of the illustrated devices is functional and conceptual, and does not necessarily have to be physically configured as shown in the figure. That is, the specific form of distribution / integration of each device is not limited to the one shown in the figure, and all or part of them may be functionally or physically distributed / physically in arbitrary units according to various loads and usage conditions. Can be integrated and configured. For example, the conversion devices 10 and 210 may be an integrated device. Further, each processing function performed by each device may be realized by a CPU and a program analyzed and executed by the CPU, or may be realized as hardware by wired logic.
　また、本実施形態において説明した各処理のうち、自動的に行われるものとして説明した処理の全部又は一部を手動的におこなうこともでき、あるいは、手動的におこなわれるものとして説明した処理の全部又は一部を公知の方法で自動的におこなうこともできる。また、本実施形態において説明した各処理は、記載の順にしたがって時系列に実行されるのみならず、処理を実行する装置の処理能力あるいは必要に応じて並列的にあるいは個別に実行されてもよい。この他、上記文書中や図面中で示した処理手順、制御手順、具体的名称、各種のデータやパラメータを含む情報については、特記する場合を除いて任意に変更することができる。 Further, among the processes described in the present embodiment, all or part of the processes described as being automatically performed can be manually performed, or the processes described as being manually performed can be performed. All or part of it can be done automatically by a known method. Further, the processes described in the present embodiment are not only executed in chronological order according to the order of description, but may also be executed in parallel or individually as required by the processing capacity of the device that executes the processes. .. In addition, the processing procedure, control procedure, specific name, and information including various data and parameters shown in the above document and drawings can be arbitrarily changed unless otherwise specified.
［プログラム］
　図５は、プログラムが実行されることにより、変換装置１０，２１０が実現されるコンピュータの一例を示す図である。コンピュータ１０００は、例えば、メモリ１０１０、ＣＰＵ１０２０を有する。また、コンピュータ１０００は、ハードディスクドライブインタフェース１０３０、ディスクドライブインタフェース１０４０、シリアルポートインタフェース１０５０、ビデオアダプタ１０６０、ネットワークインタフェース１０７０を有する。これらの各部は、バス１０８０によって接続される。 [program]
FIG. 5 is a diagram showing an example of a computer in which the conversion devices 10 and 210 are realized by executing the program. The computer 1000 has, for example, a memory 1010 and a CPU 1020. The computer 1000 also has a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, and a network interface 1070. Each of these parts is connected by a bus 1080.
　メモリ１０１０は、ＲＯＭ１０１１及びＲＡＭ１０１２を含む。ＲＯＭ１０１１は、例えば、ＢＩＯＳ（Basic　Input　Output　System）等のブートプログラムを記憶する。ハードディスクドライブインタフェース１０３０は、ハードディスクドライブ１０３１に接続される。ディスクドライブインタフェース１０４０は、ディスクドライブ１０４１に接続される。例えば磁気ディスクや光ディスク等の着脱可能な記憶媒体が、ディスクドライブ１０４１に挿入される。シリアルポートインタフェース１０５０は、例えばマウス１１１０、キーボード１１２０に接続される。ビデオアダプタ１０６０は、例えばディスプレイ１１３０に接続される。 Memory 1010 includes ROM 1011 and RAM 1012. The ROM 1011 stores, for example, a boot program such as a BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to the hard disk drive 1031. The disk drive interface 1040 is connected to the disk drive 1041. For example, a removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1041. The serial port interface 1050 is connected to, for example, a mouse 1110 and a keyboard 1120. The video adapter 1060 is connected to, for example, the display 1130.
　ハードディスクドライブ１０３１は、例えば、ＯＳ１０９１、アプリケーションプログラム１０９２、プログラムモジュール１０９３、プログラムデータ１０９４を記憶する。すなわち、変換装置１０，２１０の各処理を規定するプログラムは、コンピュータ１０００により実行可能なコードが記述されたプログラムモジュール１０９３として実装される。プログラムモジュール１０９３は、例えばハードディスクドライブ１０３１に記憶される。例えば、変換装置１０，２１０における機能構成と同様の処理を実行するためのプログラムモジュール１０９３が、ハードディスクドライブ１０３１に記憶される。なお、ハードディスクドライブ１０３１は、ＳＳＤ（Solid　State　Drive）により代替されてもよい。 The hard disk drive 1031 stores, for example, the OS 1091, the application program 1092, the program module 1093, and the program data 1094. That is, the program that defines each process of the conversion devices 10 and 210 is implemented as a program module 1093 in which a code that can be executed by the computer 1000 is described. The program module 1093 is stored in, for example, the hard disk drive 1031. For example, the program module 1093 for executing the same processing as the functional configuration in the conversion devices 10 and 210 is stored in the hard disk drive 1031. The hard disk drive 1031 may be replaced by an SSD (Solid State Drive).
　また、上述した実施形態の処理で用いられる設定データは、プログラムデータ１０９４として、例えばメモリ１０１０やハードディスクドライブ１０３１に記憶される。そして、ＣＰＵ１０２０が、メモリ１０１０やハードディスクドライブ１０３１に記憶されたプログラムモジュール１０９３やプログラムデータ１０９４を必要に応じてＲＡＭ１０１２に読み出して実行する。 Further, the setting data used in the processing of the above-described embodiment is stored as program data 1094 in, for example, the memory 1010 or the hard disk drive 1031. Then, the CPU 1020 reads the program module 1093 and the program data 1094 stored in the memory 1010 and the hard disk drive 1031 into the RAM 1012 and executes them as needed.
　なお、プログラムモジュール１０９３やプログラムデータ１０９４は、ハードディスクドライブ１０３１に記憶される場合に限らず、例えば着脱可能な記憶媒体に記憶され、ディスクドライブ１０４１等を介してＣＰＵ１０２０によって読み出されてもよい。あるいは、プログラムモジュール１０９３及びプログラムデータ１０９４は、ネットワーク（ＬＡＮ（Local　Area　Network）、ＷＡＮ（Wide　Area　Network）等）を介して接続された他のコンピュータに記憶されてもよい。そして、プログラムモジュール１０９３及びプログラムデータ１０９４は、他のコンピュータから、ネットワークインタフェース１０７０を介してＣＰＵ１０２０によって読み出されてもよい。また、変換装置１０，２１０及び学習装置２０，２２０，３２０，４２０において使用するニューラルネットワークの処理は、ＧＰＵを用いて実行される場合もある。 The program module 1093 and the program data 1094 are not limited to those stored in the hard disk drive 1031 but may be stored in, for example, a removable storage medium and read by the CPU 1020 via the disk drive 1041 or the like. Alternatively, the program module 1093 and the program data 1094 may be stored in another computer connected via a network (LAN (Local Area Network), WAN (Wide Area Network), etc.). Then, the program module 1093 and the program data 1094 may be read from another computer by the CPU 1020 via the network interface 1070. Further, the processing of the neural network used in the conversion devices 10, 210 and the learning devices 20, 220, 320, 420 may be executed by using the GPU.
　以上、本発明者によってなされた発明を適用した実施形態について説明したが、本実施形態による本発明の開示の一部をなす記述及び図面により本発明は限定されることはない。すなわち、本実施形態に基づいて当業者等によりなされる他の実施形態、実施例及び運用技術等は全て本発明の範疇に含まれる。 Although the embodiment to which the invention made by the present inventor is applied has been described above, the present invention is not limited by the description and the drawings which form a part of the disclosure of the present invention according to the present embodiment. That is, other embodiments, examples, operational techniques, and the like made by those skilled in the art based on the present embodiment are all included in the scope of the present invention.
　１０，２１０　変換装置
　１１　評価部
　１２，２１２　変換部 10,210 Conversion device 11 Evaluation unit 12,212 Conversion unit

Claims (7)

  1. 　入力された音声信号から、人が感じる音声の内容の伝わりやすさを数値化した主観評価値のいずれの値を取るかを推定する評価部と、
    　前記評価部によって推定された主観評価値を基に、所定の値の主観評価値となるように、前記入力された音声信号を変換する変換部と、
    　を有することを特徴とする変換装置。     From the input audio signal, an evaluation unit that estimates which of the subjective evaluation values that quantifies the ease with which the content of the voice felt by a person is transmitted, and the evaluation unit.
    Based on the subjective evaluation value estimated by the evaluation unit, a conversion unit that converts the input audio signal so as to have a predetermined subjective evaluation value, and a conversion unit.
    A converter characterized by having.
  2. 　前記評価部は、学習用の音声信号の特徴量と前記学習用の音声信号に対応する主観評価値との関係性を学習した評価モデルを用いて、入力された音声信号の特徴量から、主観評価情報を推定することを特徴とする請求項１に記載の変換装置。 The evaluation unit is subjective from the feature amount of the input voice signal by using the evaluation model which learned the relationship between the feature amount of the voice signal for learning and the subjective evaluation value corresponding to the voice signal for learning. The conversion device according to claim 1, wherein the evaluation information is estimated.
  3. 　前記変換部は、第１の前記主観評価値が付与された学習用の音声信号と、前記第１の主観評価値と異なる値である第２の前記主観評価値が付与された学習用の音声信号とを基に、前記第１の主観評価値と前記第２の主観評価値との違いによる音声信号の特徴量の変換を学習した変換モデルを用いて、入力された音声信号を、所定の値の主観評価値である音声信号へ変換する請求項１または２に記載の変換装置。 The conversion unit includes a learning voice signal to which the first subjective evaluation value is given and a learning voice to which the second subjective evaluation value, which is a value different from the first subjective evaluation value, is given. A predetermined audio signal is input by using a conversion model learned from the conversion of the feature amount of the audio signal based on the difference between the first subjective evaluation value and the second subjective evaluation value based on the signal. The conversion device according to claim 1 or 2, which converts a value into an audio signal which is a subjective evaluation value.
  4. 　前記評価部によって推定された主観評価値が、目標とする前記主観評価値となるように、前記入力された音声信号を変換することを特徴とする請求項１～３のいずれか一つに記載の変換装置。 The invention according to any one of claims 1 to 3, wherein the input audio signal is converted so that the subjective evaluation value estimated by the evaluation unit becomes the target subjective evaluation value. Conversion device.
  5. 　前記主観評価値は、理解のしやすさ、音声の自然さ、内容の分かりやすさ、間の取り方の適切さ、話し方のうまさ、または、印象度合いを、それぞれ数値で示したものであることを特徴とする請求項１～４のいずれか一つに記載の変換装置。 The subjective evaluation value shall be a numerical value indicating the ease of understanding, the naturalness of the voice, the ease of understanding the content, the appropriateness of how to take a pause, the goodness of speaking, or the degree of impression. The conversion device according to any one of claims 1 to 4.
  6. 　変換装置が実行する変換方法であって、
    　入力された音声信号から、人が感じる音声の内容の伝わりやすさを数値化した主観評価値のいずれの値を取るかを推定する工程と、
    　前記推定する工程において推定された主観評価値を基に、所定の値の主観評価値となるように、前記入力された音声信号を変換する工程と、
    　を含んだことを特徴とする変換方法。     It is a conversion method performed by the conversion device.
    From the input voice signal, the process of estimating which of the subjective evaluation values that quantifies the ease of transmission of the voice content that a person feels, and
    Based on the subjective evaluation value estimated in the estimation step, the step of converting the input audio signal so as to obtain the subjective evaluation value of a predetermined value, and the step of converting the input audio signal.
    A conversion method characterized by including.
  7. 　入力された音声信号から、人が感じる音声の内容の伝わりやすさを数値化した主観評価値のいずれの値を取るかを推定するステップと、
    　前記推定するステップにおいて推定された主観評価値を基に、所定の値の主観評価値となるように、前記入力された音声信号を変換するステップと、
    　をコンピュータに実行させるための変換プログラム。     From the input audio signal, the step of estimating which of the subjective evaluation values that quantifies the ease of transmission of the audio content that a person feels, and
    Based on the subjective evaluation value estimated in the estimation step, the step of converting the input audio signal so as to have a predetermined subjective evaluation value, and the step of converting the input audio signal.
    A conversion program to make a computer execute.
PCT/JP2020/042528 2020-11-13 2020-11-13 Conversion device, conversion method, and conversion program WO2022102105A1 (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
PCT/JP2020/042528 WO2022102105A1 (en) 2020-11-13 2020-11-13 Conversion device, conversion method, and conversion program
US18/036,598 US20240013798A1 (en) 2020-11-13 2020-11-13 Conversion device, conversion method, and conversion program
JP2022561229A JPWO2022102105A1 (en) 2020-11-13 2020-11-13

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/JP2020/042528 WO2022102105A1 (en) 2020-11-13 2020-11-13 Conversion device, conversion method, and conversion program

Publications (1)

Publication Number Publication Date
WO2022102105A1 true WO2022102105A1 (en) 2022-05-19

Family

ID=81602153

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2020/042528 WO2022102105A1 (en) 2020-11-13 2020-11-13 Conversion device, conversion method, and conversion program

Country Status (3)

Country Link
US (1) US20240013798A1 (en)
JP (1) JPWO2022102105A1 (en)
WO (1) WO2022102105A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025032632A1 (en) * 2023-08-04 2025-02-13 日本電信電話株式会社 Voice subjective evaluation value estimation device and voice subjective evaluation value estimation method

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH02223983A (en) * 1989-02-27 1990-09-06 Toshiba Corp Presentation support system
JPH05197390A (en) * 1992-01-20 1993-08-06 Seiko Epson Corp Speech recognition device
JP2008256942A (en) * 2007-04-04 2008-10-23 Toshiba Corp Data comparison apparatus of speech synthesis database and data comparison method of speech synthesis database
US20110295604A1 (en) * 2001-11-19 2011-12-01 At&T Intellectual Property Ii, L.P. System and method for automatic verification of the understandability of speech
JP2015197621A (en) * 2014-04-02 2015-11-09 日本電信電話株式会社 Speaking manner evaluation device, speaking manner evaluation method, and program
WO2016039465A1 (en) * 2014-09-12 2016-03-17 ヤマハ株式会社 Acoustic analysis device
US20180012613A1 (en) * 2016-07-11 2018-01-11 The Chinese University Of Hong Kong Phonetic posteriorgrams for many-to-one voice conversion

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH02223983A (en) * 1989-02-27 1990-09-06 Toshiba Corp Presentation support system
JPH05197390A (en) * 1992-01-20 1993-08-06 Seiko Epson Corp Speech recognition device
US20110295604A1 (en) * 2001-11-19 2011-12-01 At&T Intellectual Property Ii, L.P. System and method for automatic verification of the understandability of speech
JP2008256942A (en) * 2007-04-04 2008-10-23 Toshiba Corp Data comparison apparatus of speech synthesis database and data comparison method of speech synthesis database
JP2015197621A (en) * 2014-04-02 2015-11-09 日本電信電話株式会社 Speaking manner evaluation device, speaking manner evaluation method, and program
WO2016039465A1 (en) * 2014-09-12 2016-03-17 ヤマハ株式会社 Acoustic analysis device
US20180012613A1 (en) * 2016-07-11 2018-01-11 The Chinese University Of Hong Kong Phonetic posteriorgrams for many-to-one voice conversion

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2025032632A1 (en) * 2023-08-04 2025-02-13 日本電信電話株式会社 Voice subjective evaluation value estimation device and voice subjective evaluation value estimation method

Also Published As

Publication number Publication date
US20240013798A1 (en) 2024-01-11
JPWO2022102105A1 (en) 2022-05-19

Similar Documents

Publication Publication Date Title
US10468024B2 (en) Information processing method and non-temporary storage medium for system to control at least one device through dialog with user
Altonji et al. Cross section and panel data estimators for nonseparable models with endogenous regressors
WO2021005891A1 (en) System, method, and program
Alamdari et al. Personalization of hearing aid compression by human-in-the-loop deep reinforcement learning
WO2022102105A1 (en) Conversion device, conversion method, and conversion program
KR102175490B1 (en) Method and apparatus for measuring depression
US20190131948A1 (en) Audio loudness control method and system based on signal analysis and deep learning
JP7054607B2 (en) Generator, generation method and generation program
US10269349B2 (en) Voice interactive device and voice interaction method
CN113192603A (en) Mental state assessment method and system based on big data
Dingemanse et al. The relation of hearing-specific patient-reported outcome measures with speech perception measures and acceptable noise levels in cochlear implant users
WO2019058479A1 (en) Knowledge acquisition device, knowledge acquisition method, and recording medium
CN108170452A (en) The growing method of robot
KR100925828B1 (en) Quantitative Derivation Method and Sound Apparatus for Sound Quality of Vehicle Sounds
CN117672243A (en) Processing method and device for conversation of seat personnel and electronic equipment
JP7347794B2 (en) Interactive information acquisition device, interactive information acquisition method, and program
JP7459931B2 (en) Stress management device, stress management method, and program
CN112652325B (en) Remote voice adjustment method based on artificial intelligence and related equipment
JP7186207B2 (en) Information processing device, information processing program and information processing system
JP2019109789A (en) Provision device, provision method, and provision program
JP2022114906A (en) psychological state management device
KR102712474B1 (en) The Method and System That Automatically Update Nursing Chatbot Dataset
JP7561101B2 (en) Information processing device, information processing method, and information processing program
JP7164793B1 (en) Speech processing system, speech processing device and speech processing method
US20230141724A1 (en) Emotion recognition system and emotion recognition method

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20961633

Country of ref document: EP

Kind code of ref document: A1

ENP Entry into the national phase

Ref document number: 2022561229

Country of ref document: JP

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: 18036598

Country of ref document: US

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20961633

Country of ref document: EP

Kind code of ref document: A1